| 摘要 |
PURPOSE:To accurately render decision for sound/silence even under environment with high ambient noise by performing the decision for the sound/silence by using the mean value of speech power in franme unit, difference between the maximum value and the minimum value, and speech feature quantity expressed by using a reflection coefficient. CONSTITUTION:Speech in frame unit is inputted to a sound interval detector 2 after being converted to the one of PCM signal format, and simultaneously, the encode parameter of a frame one before the frame inputted to the sound interval detector 2 at that time is inputted from a high efficient speech encoder 1. The decision for the sound/silence can be performed by using the magnitude of the speech power of the present frame, the difference between the maximum value and the minimum value extending over the number of past several frames of the speech power, a value derived by using the reflection coefficient that is one of speech encode parameters obtained according to the making into high efficiency, and the code book index of a vector code obtained when vector quantization and encoding in frame unit are performed as the parameter. |
